The ML-MDLText is a groundbreaking multilabel text classifier designed to address real-world classification challenges efficiently. Built on the minimum description length (MDL) principle, ML-MDLText excels at multilabel classification without requiring problem transformation, while leveraging dependencies among labels for better accuracy. Its incremental learning capability makes it particularly suited for dynamic datasets and online applications.
Key Features
- Multilabel Classification: Handles multiple labels simultaneously without problem transformation.
- Incremental Learning: Supports dynamic learning, making it ideal for evolving datasets.
- Dependency Handling: Incorporates label dependency information for improved classification performance.
- Efficiency: Competitive results compared to state-of-the-art methods, with reduced computational costs.
Publications
- ML-MDLText: an efficient and lightweight multilabel text classifier with incremental learning
M.M. BITTENCOURT, R.M. SILVA, T.A. ALMEIDA.
Applied Soft Computing, Elsevier, Volume 96, 1-15, 2020 [pdf]
- ML-MDLText: A Multilabel Text Categorization Technique with Incremental Learning
M.M. BITTENCOURT, R.M. SILVA, T.A. ALMEIDA
Proceedings of the 8th Brazilian Conference on Intelligent Systems (BRACIS’19), 580-585, Salvador, Brazil, October, 2019 [pdf]
Code
The source code and documentation are publicly available in the ML-MDLText repository.